I N S T I T U T E

Female Empowerment and Economic Growth

Sirianne DahlumCarl Henrik KnutsenValeriya Mechkova

Working Paper SERIES 2020:103

THE VARIETIES OF DEMOCRACY INSTITUTE

June 2020

Varieties of Democracy (V-Dem) is a new approach to conceptualization and measurement of democracy. The headquarters – the V-Dem Institute – is based at the University of Gothenburg with 19 staff. The project includes a worldwide team with six Principal Investigators, 14 Project Managers, 30 Regional Managers, 170 Country Coordinators, Research Assistants, and 3,000 Country Experts. The V-Dem project is one of the largest ever social science research-oriented data collection programs.

Please address comments and/or queries for information to:

V-Dem Institute

Department of Political Science

University of Gothenburg

Sprängkullsgatan 19, PO Box 711

SE 40530 Gothenburg

Sweden

E-mail: contact@v-dem.net

V-Dem Working Papers are available in electronic format at www.v-dem.net.

Copyright ©2020 by authors. All rights reserved.

Female Empowerment and Economic Growth*

Sirianne Dahlum, PRIOCarl Henrik Knutsen, University of Oslo

Valeriya Mechkova, University of Gothenburg

June 2020

Abstract

We discuss how inclusive institutions enhance technological change, the main driver of long-term economic growth. Specifically, institutions that promote female political empowerment advance technological change through a) increasing the number and variability of new ideas introduced in the economy and b) improving the selection of more efficient ideas. We test different implications from our argument by measuring three aspects of empowerment – descriptive representation, civil liberties protection, and civil society participation – across 182 countries and 221 years. Empowerment is positively related to subsequent growth, even when accounting for initial differences in income, past growth rates, democracy, and country- and year-fixed effects. The three sub-components of empowerment are also, individually, related to growth, although not as strongly as the aggregated concept. The relationship is retained across different regimes, time periods, and geographic contexts, but is clearer for “Non-Western” coun-tries. Finally, empowerment enhances TFP growth, a proxy for technological change.

∗ The authors would like to thank Amanda Edgell and Sebastian Hellmeier as well as participants inthe V-DEM Research Seminar at the University of Gothenburg for very helpful comments and suggestions.The paper draws, in part, on earlier work funded by USAID. This research project was supported by theSwedish Research Council, Grant 439-2014-38, PI: Pam Fredman, Vice-Chancellor, University of Gothenburg,Sweden.

1

1 Introduction

To what extent does a country’s economic development rely on its political institutions? A

large literature spanning economic history, economics, and political science has been pre-

occupied with this broad, and very important, question for decades (North, 1990; Rodrik,

Subramanian and Trebbi, 2004; Gerring et al., 2005; Acemoglu and Robinson, 2012). Yet, we

lack a clear understanding of which specific institutions are more and less important, despite

researchers acknowledging that “good institutions” enhance development. A second litera-

ture emphasizes the role of inclusion of more particular social groups in positions of power

and decision-making, with a special focus on the inclusion of women. Female empowerment

is not only a normative ideal in itself, but may have instrumental value for other valuable

outcomes (e.g., Sundstrom et al., 2017), including economic development (e.g., Duflo, 2012).

In this paper, we bridge these two literatures by focusing on how open and inclusive

political institutions influence countries’ trajectories of economic development by empow-

ering and including a broad population group that is otherwise often excluded, namely

women. We rely on a broad definition of female political empowerment, which includes in-

creased capacity for women to influence political decision-making through three pathways:

1) descriptive political representation; 2) freedom of choice, guaranteed by protected civil

liberties; and 3) opportunity to express their voice. Existing studies on political institutions

and development have mainly focused on how institutions influence capital investments. Yet,

growth economists propose that technological change is the main driver of long-term growth

([ Romer, 1990; Acemoglu, 2008). We present an argument and empirical analysis indicating

how female empowerment contributes to technological change. Specifically, our theoretical

argument focuses on how institutional features that promote female political empowerment

affect technological change through a) increasing the number and variability of new ideas

introduced in the economy and b) improving how efficiently the best, new ideas are adopted.

Women constitute the majority of the adult population in many countries, and excluding

women from processes of idea generation and selection should have substantial implications

1

for a society’s ability to generate technological change.

We test various implications from this argument using extensive data from the Varieties

of Democracy (V-Dem) project (Coppedge et al., 2020) to measure the three mentioned

aspects of female political empowerment. Across 182 countries and time series extending for

221 years, we find robust evidence that female political empowerment (FPE) is positively

related to subsequent GDP per capita growth. This relationship holds up when accounting

for initial differences in economic development, democracy levels, and country- and year-

fixed effects. When disaggregating FPE into its sub-components, we find that descriptive

political representation, civil liberties protection, and civil society participation are all, in-

dividually, related to growth. Further, the overall relationship between FPE and growth is

retained across different contexts, but is stronger and more robust for “Non-Western” than

for “Western” countries. Finally, when disaggregating the sources of economic growth, we

find that FPE enhances total factor productivity growth, a proxy for technological change.

In the following, we first review relevant studies on institutions and economic growth,

before we consider studies on the consequences of political inclusion and representation

of women, more specifically. Next, we present our theoretical argument on how female

political empowerment enhances technological change, which, in turn, enhances economic

growth. We thereafter present our data and research design, before we present and discuss

our empirical results. In the concluding section, we discuss the real-world relevance of our

findings. For many people, including political leaders, female political empowerment is of

intrinsic normative value, and additional motivation for ensuring equal participation and

protection of rights across genders is not needed. Insofar as women’s rights are human

rights (Bunch, 1990), women should have the same basic opportunities as men, including

an equal say in decisions on how to govern society. Yet, countries across the world still

vary enormously in how empowered women are, politically. The “business case” that we

present might contribute to incentivizing initially hesitant leaders and social groups – albeit

for instrumental reasons – to improve female political empowerment.

2

2 Relevant literatures

In this section, we provide a brief overview of the theoretical and empirical literatures that

serve as building blocks for our argument, which is presented in the following section. We

first review studies on (immediate and deep) determinants of economic growth, focusing on

arguments and evidence pertaining to how institutions shape technological change. Next, we

review studies addressing how different aspects of female empowerment influence economic

outcomes.

2.1 Economic growth, and the role of institutions

Growth economists have, for decades, studied the “immediate determinants” of growth (e.g.,

Barro and Sala-i Martin, 2004; Helpman, 2004; Acemoglu, 2008). Several theoretical mod-

els specify how different such determinants feed into growth processes (e.g., Solow, 1956;

Mankiw, Romer and Weil, 1992; Romer, 1990) and “growth accounting” exercises (e.g.,

Young, 1995; Klenow and Rodriguez-Clare, 1997; Baier, Dwyer Jr. and Tamura, 2006) have

assessed how much of growth in national income or production (both typically measured by

GDP per capita) come from the various determinants. Immediate determinants are either

classified as factor inputs in production processes – notably labour hours, physical capital,

human capital, land, and natural resources – or as ways in which these inputs are combined

into producing output, referred to by the broad concept of “technology”. This concept covers

specific production technologies, but also ideas about economic policies and how economic

processes are organized. The presumed relative importance of different immediate determi-

nants in influencing short-, medium-, and long-term growth varies across theoretical models.

Yet, the most prominent ones – both among so-called neo-classical- and endogenous growth

models – highlight that accumulation of factor inputs, such as labor and capital, may boost

growth in the short- to medium term, but not in the longer term (as returns to accumulating

3

more inputs decrease). In contrast, technological change drives long-run growth.1

The introduction of new ideas and production technologies to an economy can come

from domestic innovation or from the adoption (and possibly adaptation) of technology

developed abroad. Several economists focus primarily on processes of innovation for un-

derstanding technological change. Romer (1990), for example, introduces a “new growth

model”, where profit-maximizing firms contribute to technological change by innovating and

supplying a wider variety of new products. Grossman and Helpman (1991) and Aghion and

Howitt (1992) model technological change as generated by firms investing in innovation of

improved products that replace existing products of inferior quality. Yet, since ideas are

“non-rivalrous” (see Romer, 1993), production and organization technologies can, at least

in principle, be used to enhance efficiency also in other countries than where they originate

from. Indeed, most production and organization technologies in use in any current econ-

omy come from abroad. Diffusion of foreign technology is especially important for small

countries and poor countries far away from the “global technological frontier”. Hence, in

order to understand technological change, and thereby persistent differences in growth rates

across countries, we must understand why some countries are better than others at adopting

production techniques and ideas developed elsewhere, and at diffusing them within their

economies.

So-called “evolutionary growth models” (see, e.g., Nelson and Winter, 1982; Nelson, 2005;

Verspagen, 2005) are relevant in this regard. This strand of growth theory has developed

models that draw on key notions from evolutionary biology to assess which factors enhance

the adoption of new and more efficient technologies. The two key inputs to such processes

are a) an increased variety of new ideas being introduced to the economy – partly from

domestic processes of innovation, but notably through diffusion of ideas from abroad – and

b) mechanisms for ensuring the selection of the more efficient ideas. A large variety of

1We remark that the sharp distinction between how factor inputs and technology feed into growth is asimplification; investments in new machinery may introduce new technology (Nelson, 2005) and high humancapital levels facilitate the adoption of more efficient technologies (Kremer, 1993a).

4

competing ideas and technologies enhances economic efficiency, especially when it is unclear

a priori how ideas and technologies will work in practice; economic actors learn how effective

they are from processes of trial and error (North, 2005). Regarding the selection of ideas,

this process reduces variety as new techniques are adopted through learning and less efficient

techniques are discarded. An economy thus requires the steady introduction of novel ideas

to keep up variety. Factors that simultaneously allow for the introduction of new ideas and

enable improved selection and diffusion processes are therefore especially likely to enhance

technological change. This insight is central in our theoretical argument below.

So-called deeper determinants of economic growth (Rodrik, Subramanian and Trebbi,

2004) are located prior in the causal chain to the immediate determinants discussed above.

Suggested deeper determinants include cultural norms and practices, various geographic

features, and demographic factors (see, respectively, Landes, 1998; Diamond, 1997; Kremer,

1993b). Yet, the perhaps most studied deeper determinant is “good institutions” (e.g., North,

1990; De Long and Shleifer, 1993; Rodrik, Subramanian and Trebbi, 2004; Acemoglu, John-

son and Robinson, 2001; Acemoglu and Robinson, 2012). By influencing which economic

policies are selected, and determining the expected costs and risks to investors, institutions

presumably affects capital accumulation (e.g., North, 1990; Bizzarro et al., 2018). But, more

importantly for long-term growth, institutions may also influence innovation and the adop-

tion of new technologies. For example, institutions ensuring the protection of intellectual

property rights may strengthen incentives for firms to invest in innovation activities (Romer,

1990). Further, protection of civil liberties (Knutsen, 2015) or competitive multi-party elec-

tions (North, Wallis and Weingast, 2009) may enhance both the variety of ideas introduced

into the economy and improve selection of the more efficient ideas. Open and inclusive po-

litical institutions “more readily generate a range of solution to problems; they more readily

experiment with solutions to problems; and they more readily discard ideas and leaders who

fail to solve them” (North, Wallis and Weingast, 2009: 134). By enabling different popu-

lation groups – and thus more creative minds – to enter political debates and take part in

5

economic interactions, open and inclusive institutions enhances technological change, and

thereby growth.

Despite the plausibility of the argument that “institutions matter” for growth, scholars

have yet to conclude on exactly which particular institutions that matter the most. Some

authors focus on features of the public administration (Evans and Rauch, 1999; Fukuyama,

2014). Others argue that low levels of corruption and impartiality determine development

outcomes (Rothstein and Teorell, 2008). Yet others highlight the role of institutionalized

political parties (Bizzarro et al., 2018). Finally, democracy, and especially competitive multi-

party elections, is another prominent explanation for economic growth, although empirical

results are not robust (Przeworski et al., 2000; Gerring et al., 2005; Acemoglu et al., 2019).

In this paper, we focus on institutions that further female political empowerment. We argue

that such institutions enhance growth through enhancing both the variety of new ideas and

the selection of more efficient ones, thereby leading to more rapid technological change.

Before we present our argument, however, we review existing studies that relate aspects of

female empowerment to economic outcomes.

2.2 Economic consequences of female empowerment

Several studies have proposed that gender equality and female empowerment may relate

to economic outcomes, including growth and its immediate determinants (for reviews, see

Cuberes and Teignier, 2014; Duflo, 2012; Kabeer and Natali, 2013). The most intensively

studied outcomes are female labor participation and education outcomes.

Regarding the former, Esteve-Volart (2004) presents a theoretical model indicating the

inefficiency and negative economic consequences of excluding women from labor participa-

tion. In this model, individuals are born with a given talent, and restricting the access of

women to managerial positions leads to loss of talent in the positions where they are the

most productive. This gives diminished innovation and slower technology adoption, thereby

reducing productivity growth. Such exclusion of women thus gives lower equilibrium wages,

6

both for male and female workers. Further, restricting the type of work women can do,

more generally in the economy, to only home production, reduces income also due to the

inherently lower productivity of such production. Finally, both types of exclusion – from

managerial positions and from general production in certain sectors – leads to lower invest-

ment in human capital, which further contributes to lower growth rates. Similarly, building

a model of heterogeneous talents in a population, Cuberes and Teignier (2012) show how

barriers for women to become managers significantly reduce the average talent available in

the economy, and thereby aggregate productivity and income levels. Their cross-country

estimates indicate that the loss in GDP per capita is about 12 percent when women cannot

take managerial positions, and about 40 percent when women are completely excluded from

the labor market. The estimated income loss in the mid-2000s for countries in the Middle

East and North Africa, the region with the highest exclusion rates for women, is 27 percent.

Similarly, studies suggest that gender gaps in education hurt economic growth directly due

to reduced human capital, with potential ramifications also for technological change (Klasen,

2002; Klasen and Lamanna, 2009; Knowles, Lorgelly and Owen, 2002; Thevenon et al., 2012).

Education for women also carries other externalities, such as reduced fertility and improved

child-care and child survival, that enhance the human capital of future generations, and

thus growth (for reviews, see Mitra, Bang and Biswas, 2015; Duflo, 2012). Using panel data,

Klasen and Lamanna (2009) and Thevenon et al. (2012) investigate the effects of gender

gaps in education and labor force participation, and find that such gaps are associated with

reduced economic growth. In OECD countries, on average, an additional year of education

for girls is estimated to give 10 percent higher GDP per capita (Thevenon et al., 2012).

In sum, the literature convincingly shows that there is a positive relationship between

higher education levels and labor force participation among women and economic growth.

However, we know less about the effects of other types of female empowerment. Mitra, Bang

and Biswas (2015) argue that gender equality is a multi-dimensional concept, consisting of

distinct features that may have different effects on economic growth. Mitra, Bang and Biswas

7

(2015) thus create two distinct indices of female empowerment. They find that equality in

economic opportunity (a combined index of a literacy gap, secondary enrollment gap and

fertility rate) is associated with growth in developing countries, while equality in economic

and political outcomes (index of labor force participation gap and percent of women in

parliament) displays this association in developed economies. Yet, despite the evidence

presented by Mitra, Bang and Biswas (2015), we still lack in understanding of exactly how

the political exclusion of women, along different dimensions, affect economic growth. In

the following, we argue that female political empowerment have positive implications for

technological change in both developing and developed countries.

3 Argument

Following (Sundstrom et al., 2017), we adopt a broad definition of female political empow-

erment (FPE) as “a process of increasing capacity for women, leading to greater choice,

agency, and participation in societal decision-making”. Hence, we go beyond descriptive

political representation and also cover freedom of choice guaranteed through civil liberties

protection and the ability to voice ideas and preferences. Rather than focusing on whether

women have access to particular resources such as education or land, we consider the extent

to which women have access to political power and are able to distribute resources and in-

fluence decisions, more generally (Longwe, 2000). Before we dig deeper into the particular

mechanisms, linked to various aspects of FPE and how they influence the variation in new

ideas or the selection of more efficient ones, we summarize the core logic of the argument.

Figure 1 gives a graphical illustration of the main steps in the argument. Our concept

of FPE has three sub-components. The first sub-component pertains to enhanced freedom

of choice for women in different spheres, notably related to strengthened civil liberties. The

second relates to improved representation for women in key arenas of political decision-

making, including the legislature and executive. The third pertains to women being able

8

Fem

ale

Em

po

wer

men

t Freedom of choice

Representation

Voice

Variation

ideas

Selection

ideas

Technological

change

GDP

p.c.

growth

Figure 1: Sketch of the main components and links in our argument

to actively voice their preferences and ideas though various forms of civic participation.

Political institutions that enhance any one of these sub-components may also enhance the

rate of technological change. As Figure 1 shows, we surmise that all three sub-components

have independent effects on the variety of new ideas introduced into the economy as well as

the selection of more efficient ideas. These are the two key determinants of technological

change, according to the evolutionary growth models reviewed above. Since technological

change shapes economic growth, we further anticipate links between all three sub-components

and GDP per capita growth rates, and an even stronger link between the aggregated concept

of FPE and growth. Let us now turn to plausible, more specific mechanisms, which we sort

according to FPE’s three sub-components.

3.1 Descriptive political representation

Arguments along the lines of Esteve-Volart (2004), which suggest that excluding women (and

thus about half the population of any country) from key positions is economically inefficient,

can be translated to the area of political representation. Political arenas such as legislatures

9

and executives (or local councils, for that matters) are where many vital decisions about how

a society develops, including its economy, are made. If we assume that a) economic and other

policies matter economic development, and b) the quality of policies depends on characteris-

tics of the decision makers, we can expect that changes in descriptive political representation

affect development.2 Bringing in women in politics not only expands the country’s “political

talent pool”, but evens increases the variance in other relevant characteristics of representa-

tives such as types of experience, knowledge, or even policy preferences (e.g., Khan, 2017).

This, in turn, enhances the quality of deliberation by bringing in new and different ideas,

and thereby increases the chances of adopting policies that benefit a broad segment of the

population (Mansbridge, 1999). Hence, improved descriptive representation of women may

increase both the variation of policy ideas and improve the process of selecting the “best”

such ideas.

Existing studies show systematic differences in the policy preferences of women and

men (Khan, 2017). Given these differences, increased female representation may lead to

the selection of certain policies that are (objectively) better at generating at least certain

development outcomes. At the micro level, women invest more in goods and services that

improve the well-being of families and that improve education and health-care outcomes

(Duflo, 2012). At a more aggregate level, Miller (2008) shows that introduction of women

suffrage in the United States was associated with declining infant mortality due to the

qualitatively different issues that women placed on the political agenda, notably related

to health-care. Elite-level analysis reveal that female candidates present themselves in a

systematically distinct manner from men in campaigns and more often promote health-care

and education issues (Kahn, 1993). These patterns are replicated in recent years and in online

2Phillips (1995), for example, highlights that the personal characteristics of representatives are relevantfor the representation of the population and their interests, with further implications for which policiesare produced. The basic premise is that descriptive representation is required to ensure that everyone’sinterests and points of view are heard and taken into account (Birnir and Waguespack, 2011). In politicalenvironments with competition for resources and agendas, who the political representatives are matter. Ifequal representation is not achieved, adopted policies will reflect the preconditions and preconceptions of thedominant group (Young, 2011).

10

behavior (Evans and Clark, 2016; Mechkova and Wilson, 2019). Swiss, Fallon and Burgos

(2012) examine how descriptive representation influences child health across 102 developing

countries from 1980-2005, and find that compared to countries with no women in parliament,

countries meeting a 20-percent threshold experience increased rates of immunizations and

infant- and child survival.

Improved descriptive representation also has symbolic significance (Pitkin, 1967), which

could, in turn, have different substantive effects. A more representative government might

be one that citizens trust more and are more likely to engage with (Mansbridge, 1999).

Female voters may be more likely to contact female political representatives, perceiving

that their interests are better defended by someone with similar background (Mechkova and

Carlitz, 2018). Indeed, female citizens more often attend village meetings and express their

points of view with women in the local leadership (Beaman et al., 2009). Such feedback and

interactions between citizens and policy makers is crucial for identifying what policies are

appropriate for the local context and for effectively implementing them, in practice, with

downstream implications for productivity (Evans, 1995).

Better political representation can also enhance female participation in various economic

arenas. Ghani, Mani and O’Connell (2013) examines mandated political representation at

the local level in India, and find that higher female representation over extended time relates

to greater female labor force participation, partly from increased public sector employment

and partly from the building of infrastructure (e.g., related to roads and health-care) that

facilitates women entering the labor force.3 And, as proposed by Esteve-Volart (2004) and

others, increased female labor force participation leads not only to a more heterogeneous

pool of workers, but also to decision-makers in the economy, on average, being more talented,

thereby enhancing productivity growth.

3Chattopadhyay and Duflo (2004) also find evidence from India that elected local leaders invest in infras-tructure that is prioritized by citizens of their own gender.

11

3.2 Freedom of choice

Civil liberties include various private and political liberties (e.g., freedom of expression and

movement), physical integrity rights (e.g., freedom from forced labor and torture), as well

as property rights and rule of law with access to impartially administered justice. Such

liberties are differentially protected across countries and political systems, but also varies

between groups within a country. Typically, women’s liberties are worse protected than

men’s (e.g., World Bank, 2020b). This lack of protection for women may have downstream

implications for macroeconomic performance. Several studies propose that the protection of

different civil liberties matter for growth through affecting incentives to invest in capital and,

notably, via influencing processes of innovation and idea diffusion (and thus technological

change). Insofar as women constitute half of the population, arguments credibly linking the

protection of civil liberties to technological change and economic growth should be highly

relevant also when it comes to women’s civil liberties, more specifically. We review two

relevant such arguments.

One prominent “institutionalist explanation” of economic growth focuses on institutions

that ensure property rights are protected for broad segments of the population (e.g., North,

1990; Acemoglu, Johnson and Robinson, 2001). Assessments of risks and the expected profits

of prospective investment objects hinge on investors’ perception of whether their future

rights to the investment object (and revenue generated from it) are protected from theft,

expropriation, and other infringements. If so, the expected returns to an investment object

more likely outweigh expected costs, leading to more investments and thus higher income

levels (Olson, 1993). Importantly for our argument focusing on technological change, well-

functioning rule of law and stable property rights reduce various risks and expected costs

of investing in costly research and development-related activities (e.g., North, 1990; Romer,

1990). Whenever poor property rights protection pertains to half the adult population

(women), aggregated investment levels and productivity growth will decline.

Adding to the general argument, Goltz, Buche and Pathak (2015) find an interaction

12

effect between rule of law and women’s descriptive representation on female entrepreneurship

– reforms aimed at stimulating women’s economic participation, enforced by female political

representatives, may be less effective when rule of law is weak. However, Goltz, Buche

and Pathak (2015) consider rule of law at the country-level without accounting for women

often facing disproportionate infringements. Goldstein and Udry (2008) study Ghana, where

women have less secure tenure rights than men. This hinders women from leaving their land

for a long fallow, despite the clear productivity benefits of this practice when fertilizers are too

expensive. The result is lower productivity on female-owned plots, and even within the same

household, women achieve significantly lower profits than their husbands (p.995). Similarly,

Duflo (2012) proposes the relatively weaker property rights for women as an explanation for

why households invest less in labor and fertilizers in plots owned by women.

The second type of argument focuses on private and political liberties – notably freedoms

of speech, media, and movement – for increasing variation in new ideas and for selecting

the more efficient ones.4 Knutsen (2015) details how free speech and open debate allow

entrepreneurs, decision-makers in firms, non-governmental organizations, bureaucrats, and

politicians to better adopt and disseminate ideas from abroad and identify and discard

less efficient solutions. Even when motivated by purely political reasons such as restricting

opposition mobilizing against the regime, restrictions on communication and free speech may

unintentionally suppress the diffusion of economically relevant ideas; in practice, it is very

difficult to enforce restrictions on free speech that filter out politically from economically

relevant ideas. In other words, different actors may more freely identify and disseminate

new organization and production techniques when civil liberties are protected. Protection of

such liberties also enables a more critical evaluation of ideas – including critical comparisons

of new ideas and technologies to old, traditional ones – thereby enhancing the selection of

more efficient ones. Knutsen (2015) finds empirical support for the notion that stronger civil

4Estrin and Mickiewicz (2011) considers the economic consequences of gender-specific violations of suchrights, and finds that violations on freedom of movement affect women disproportionately, with negativeconsequences for female employment.

13

liberties protection enhance technological change and, subsequently, economic growth (see

also Dahlum, Knutsen and Lindberg, 2018). This leads us to expect that stronger protection

of civil liberties for women, more specifically, enhances technological change and economic

growth.

3.3 Voice

Finally, we consider whether ordinary women are able to effectively voice their preferences

and ideas through civic participation, be it through political discussions in the private sphere

or through various organizations. As summarized by Sundstrom et al. (2017), to be politi-

cally empowered, women must have the opportunity to freely express political views, organize

collectively, and be represented in the ranks of journalists. As already indicated in our dis-

cussion of civil liberties, the openness of societies aids the adoption of new and more effective

technologies, as various freedoms of discussion, press, and organization promote the learn-

ing and critical assessment of new ideas. Further, in closed societies policy-development

and innovation might be hindered if people are restricted from partaking in organizations

and engaging in other forms of collective action. Non-governmental organizations – due to

their specialized knowledge and by voicing the preferences of relevant, interested parties –

play a prominent role in providing inputs to the formulation and effective implementation of

policies (Evans, 1995). Restricting the ability of key population groups to organize and ac-

tively partake in civil society thus restricts relevant feedback to the government officials who

formulate economic policies. In societies where civil society participation and information

sharing between non-governmental organizations and the government is heavily regulated or

even forbidden, fewer, unconventional inputs and viewpoints are presented to policy makers,

making it harder for them to identify the full range of options or detect flaws in favored

policies (North, 2005).

Women being included in political discussions and organizational life should thus improve

the quality of policies by expanding the variety of inputs to policy-making processes (Birnir

14

and Waguespack, 2011). Further, contexts with representation of diverse interests may also

produce a more cooperative atmosphere, where minority groups are more likely to speak out

to defend their interests and the dominant group more prepared to listen to different views

(Kanter, 2008). Thus, in gender-inclusive organizations and societies, new and alternative

viewpoints on economic policies may be stimulated, helping policy makers in the inherently

difficult task of selecting efficient policies with potential macroeconomic benefits.

4 Data

Until recently, data limitations would have hindered our ability to systematically test im-

plications from the above argument on extensive data material. However, the recent V-

Dem dataset, v.9 (Coppedge, Gerring, Knutsen, Lindberg, Teorell, Altman, Bernhard,

Fish, Glynn, Hicken, Luhrmann, Marquardt, McMann, Pemstein, Seim, Sigman, Skaaning,

Staton, Wilson, Cornell, Gastaldi, Gjerlow, Ilchenko, Krusell, Maxwell, Mechkova, Medzi-

horsky, Pernes, von Romer, Stepanova, Sundstrom, Tzelgov, Wang, Wig and Ziblatt, 2019;

Coppedge, Gerring, Knutsen, Lindberg, Teorell, Altman, Bernhard, Fish, Glynn, Hicken,

Luhrmann, Marquardt, McMann, Paxton, Pemstein, Seim, Sigman, Skaaning, Staton, Cor-

nell, Gastaldi, Gjerlow, Mechkova, von Romer, Sundstrom, Tzelgov, Wang, Wig and Ziblatt,

2019) contains measures are well suited for the purpose. The measures that we employ have

extensive coverage and match up well with the theoretical concepts of interest by coding

gender-specific features of political representation, civil liberties, and civil society participa-

tion. Hence, we can capture the different, relevant aspects of female political empowerment

while simultaneously conducting stringent tests that require long time series, for example by

including country- and year-fixed effects in our models.

We refer to Coppedge et al. (2020) for details on the construction, methodology and con-

tents of the V-Dem dataset. But, in brief, the dataset is constructed to ensure measures that

are comparable across countries and over time, and that carry a high degree of reliability

15

and validity. The data-generating process and aggregation schemes for the different indica-

tors and indices are fully transparent. About half of the indicators are more objective and

coded by research assistants (e.g., share of adult population with de jure voting rights) and

the other half are more evaluative in nature (e.g., extent of election violence) and assigned

scores on the basis of expert surveys. Normally, at least five independent experts score each

indicator (per country-year). Experts vary by question/subject area and country, and are

recruited based on their documented expertise in the particular area. Thus, the raw data

come from more than 3,200 experts, in total. V-Dem combines the assessments from different

experts by using a Bayesian item response measurement model that takes into account each

expert’s reliability, determined, inter alia, by level of agreement with other country experts

(for details, see Pemstein et al., 2018; Coppedge et al., 2020).

Concerning our main independent variable, we follow Sundstrom et al. (2017) in defin-

ing female political empowerment as a “a process of increasing capacity for women, leading

to greater choice, agency, and participation in societal decision-making”. We measure this

concept by drawing on V-Dem’s Female Political Empowerment index (FPE). This index

consists of three sub-indices, which are equally-weighted in the aggregation of the overall

index by taking the simple average. The first sub-index is the women’s civil liberties index,

which largely captures our theoretical freedom of choice sub-component. It is formed by

taking the point estimate from a Bayesian factor analysis on four expert-coded items. The

second sub-index is the women’s civil society participation index, which is a latent factor vari-

able estimated on three items and roughly corresponds to the theorized voice sub-component

of female political empowerment. The final sub-index is the women’s political participation

index, which captures the representation sub-component and is constructed by averaging two

indicators. Table 1 lists all indicators included in each of the three sub-indices. The aggre-

gated FPE ranges from, 0–1, where 1 indicates high level of female political empowerment.

Our main dependent variable is GDP per capita growth, measured in annualized, per-

centage terms. We mainly draw on estimates of Ln GDP per capita from Fariss et al. (2017),

16

Table 1: Components and indicators entering V-Dem’s Women Political Empowerment Index

Female Political Empowerment Index

Women civil liberties index

Freedom of domestic movement womenFreedom from forced labor womenProperty rights womenAccess to justice women

Women civil society participation indexFreedom of discussion womenCSO women’s participationPercent female journalists

Women political participation indexPower distributed by genderLower chamber female legislators

but also run tests employing GDP data from the Maddison project (Bolt and van Zanden,

2013). The former data source allows us to extend the analysis back to 1789 and include 182

polities in our benchmark regression, whereas the latter extend back to 1800 and allow us to

include 163 polities. The estimates on (ln) GDP per capita data from Fariss et al. are arrived

at by using a dynamic latent trait model and drawing on information from different, existing

GDP and population datasets, including the Maddison data.5 One benefit with Fariss et

al.’s latent model estimation routine is that it mitigates various kinds of measurement error.

These data also mitigate missing values by imputation. For tests conducted on the original

Maddison series, we interpolate these data – which are often measured every tenth year in

the 1800s – by assuming constant growth rates across intervals with missing data. Since the

Fariss et al. time series are imputed, and predictions are presumably poorer for observations

without scores even on the extensive Maddison series, many error-prone observations are

likely dropped when using the original Maddison series. In sum, the two GDP sources have

different validity and reliability issues, and should complement each other well.

Our second dependent variable pertains to technological change. While researchers have

aimed to capture technological change with several indices and proxies (see, e.g., Knutsen,

2015), most measures lack extensive time series or cross-country coverage. The most com-

monly used proxy among growth economists is growth in Total Factor Productivity (TFP).

5Indeed, we use the version of the Fariss et al. estimates that are benchmarked in the Maddison timeseries.

17

TFP growth is basically calculated as residual economic growth after removing growth stem-

ming from changes in physical capital, human capital, and labor supply.6 We draw on the

extensive TFP data from Baier, Dwyer Jr. and Tamura (2006), which cover 145 countries

with several time series extending back to the 19th century – the earliest measurement is the

United Kingdom in 1831. Baier, Dwyer Jr. and Tamura (2006) draw on various sources to

produce their growth accounting estimates, notably the Penn World Tables, World Devel-

opment Indicators, the Maddison project, and Mitchell’s historical statistics (for details, see

Baier, Dwyer Jr. and Tamura, 2002: pp. 24–26). Given the paucity of relevant historical

data sources, Baier et al. only calculate TFP with uneven intervals, and with years of mea-

surement differing across countries. Typically, the time series include a data point for about

every tenth year. We therefore follow the approach in Knutsen (2015), and interpolate these

time series by assuming constant annual growth rates in TFP in between two observations.

In the Appendix, we present descriptive statistics and map distributions of the main

variables discussed above. Regarding data sources and measures for the control variables,

we introduce them in the next section when discussing our different regression specifications.

5 Empirical analysis

We start out by assessing the empirical implication that countries where women are empow-

ered politically experience more rapid economic growth. Next, we detail this relationship by

considering whether it applies to different geographical and temporal contexts, but also by

looking more closely into whether particular sub-components of female empowerment drive

the results. Finally, we investigate the relationship between FPE and TFP growth.

18

Japan

Burma/Myanmar

Egypt

Argentina

India

South Korea

ThailandIndonesia

Nepal

Germany

Iran Morocco

Tunisia

Turkey

China

Austria

Cuba

Denmark

66.

57

7.5

8Ln

GD

P pe

r cap

ita

0 .1 .2 .3 .4 .5Female Political Empowerment

Mexico

Sweden

Switzerland

South AfricaJapan

Burma/Myanmar

Egypt

Colombia

Brazil

United States of America

Portugal

Bolivia

Peru

Argentina

India

South Korea

ThailandVenezuela

Indonesia

Nepal

Canada

Chile

Ecuador

FranceGermany

Iran

Italy

Morocco

NetherlandsSpain

Tunisia

Turkey

United Kingdom

Uruguay

ChinaJamaica

Sri Lanka

Belgium

Bulgaria

Cuba

Denmark

FinlandGreece

New Zealand

Norway

RomaniaSaudi Arabia

Serbia

Singapore

67

89

Ln G

DP

per c

apita

0 .2 .4 .6 .8Female Political Empowerment

Mexico

SwedenSwitzerland

Ghana

South Africa

Japan

Burma/Myanmar

AlbaniaEgypt

Yemen

Colombia Poland

Brazil

United States of America

Portugal

El SalvadorBolivia

Haiti

Honduras

Mali

PeruSenegal

Vietnam

Afghanistan

Argentina

Ethiopia

IndiaKenya

North Korea

South Korea

Lebanon

PhilippinesTaiwan

ThailandUganda

Venezuela

BeninBurkina Faso

Cambodia

Indonesia

MozambiqueNepal

Nicaragua

NigerZambia

Zimbabwe

Guinea

Ivory Coast

Mauritania

CanadaAustralia

Botswana

Burundi

Cape VerdeCentral African Republic

Chile

Costa RicaEcuador

FranceGermany

GuatemalaIran Iraq

IrelandItaly

Jordan

Lesotho

LiberiaMalawi

Mongolia

Morocco

Netherlands

Panama

Spain

Syria

TunisiaTurkey

United Kingdom

Uruguay

Chad

China

Democratic Republic of the Congo

Republic of the Congo

Djibouti

Dominican Republic

Gabon

The Gambia

Guinea-BissauLaos

Madagascar

Namibia

Rwanda

Sri Lanka

SwazilandTogo

AustriaBarbados

Belgium

Bulgaria

Comoros

CubaCyprus

Czech Republic

Denmark

Equatorial Guinea

Finland

Greece

Hong Kong

Iceland

Israel

Luxembourg

MalaysiaMauritius

New Zealand

Norway

Oman ParaguayRomania

Sao Tome and PrincipeSerbia

Seychelles

Singapore Hungary

67

89

10Ln

GD

P pe

r cap

ita

0 .2 .4 .6 .8Female Political Empowerment

Mexico

SwedenSwitzerland

Ghana

South Africa

Japan

Burma/Myanmar

Russia

AlbaniaEgypt

Yemen

Colombia

PolandBrazil

United States of America

Portugal

El Salvador

Bangladesh

BoliviaHaiti

Honduras

Mali

Pakistan

Peru

SenegalVietnam

Afghanistan

Argentina

Ethiopia

IndiaKenyaNorth Korea

South Korea

Lebanon

NigeriaPhilippines

Tanzania

Taiwan

Thailand

Uganda

Venezuela

BeninBurkina FasoCambodia

Indonesia

Mozambique

Nepal

Nicaragua

Niger

Zambia

Zimbabwe

Guinea

Ivory CoastMauritania

CanadaAustralia

Botswana

Burundi

Cape Verde

Central African Republic

Chile Costa Rica

Ecuador

FranceGermany

GuatemalaIran

Iraq

IrelandItaly

Jordan

Latvia

Lesotho

Liberia

Malawi

Mongolia

Morocco

Netherlands

Panama

Sierra Leone

Spain

Syria

TunisiaTurkey

Ukraine

United Kingdom

UruguayAlgeria

Angola

ArmeniaAzerbaijan

Belarus

Cameroon

Chad

China

Democratic Republic of the Congo

Republic of the CongoDjibouti

Dominican RepublicGabon

The Gambia

Georgia

Guinea-Bissau

JamaicaKazakhstan

KyrgyzstanLaos

Libya

Madagascar

Moldova

NamibiaPalestine/West Bank

Rwanda

Sri Lanka

Swaziland

Tajikistan Togo

Trinidad and Tobago

TurkmenistanUzbekistan

AustriaBahrain

BarbadosBelgium

Bosnia and HerzegovinaBulgaria

Comoros

Croatia

Cuba

CyprusCzech Republic

Denmark

Equatorial Guinea Estonia

FinlandGreece

Hong KongIcelandIsraelKuwait

Lithuania

Luxembourg

MacedoniaMalaysia

MaltaMauritius

New Zealand

Norway

Paraguay

Romania

Sao Tome and Principe

Serbia

Seychelles

Singapore

SlovakiaSlovenia

United Arab Emirates

Hungary

68

1012

Ln G

DP

per c

apita

0 .2 .4 .6 .8 1Female Political Empowerment

Figure 2: Scatter-plots, overlaid with (bivariate) best-fit lines and 95% confidence inter-vals, for Female Political Empowerment (data taken from V-Dem; x-axes) and Ln GDPper capita(data taken from the Maddison project; y-axes) in the years 1830 (top-left), 1900(top-right), 1950 (bottom-left) and 2000 (bottom-right).

5.1 Main analysis: Female empowerment and economic growth

Before we present our benchmark panel regression, we consider some descriptive statistics and

cross-country correlations. The scatterplots in Figure 2 illustrate the positive cross-country

correlation that have existed – and been fairly persistent through modern history – between

our Female Political Empowerment (FPE) index and Ln GDP per capita as measured by

the Maddison project. Figure 3 shows equivalent plots based on the Fariss et al. data. More

specifically, the panels display scores and the best linear fit from the years 1830, 1900, 1950,

and 2000. Also for (annual) GDP per capita growth rates, there is a clear difference, on

6Since it is calculated as residual growth after, TFP growth can stem from other processes than techno-logical change that are left unaccounted for in the growth accounting exercise. These include increases inprices for major exports and natural resource discoveries. Yet, technological change is widely considered asthe main source behind TFP growth, especially in the longer run, by growth economists.

19

Switzerland

JapanBurma/Myanmar

Russia

Egypt

Yemen

Afghanistan Argentina

Ethiopia

India

South Korea

ThailandIndonesia

Nepal

IranMoroccoTunisia

Turkey

China

Madagascar

Austria Cuba

Denmark

Paraguay

Saudi Arabia

Modena Parma

Tuscany

Two Sicilies

Papal States

56

78

9Ln

GD

P pe

r cap

ita

0 .1 .2 .3 .4 .5Female Political Empowerment

Mexico

Sweden

Switzerland

South Africa Japan

Burma/Myanmar

Russia

EgyptColombiaBrazil

United States of America

PortugalEl Salvador

BoliviaHaiti

Honduras

PeruSudanAfghanistan

Argentina

Ethiopia

India

South Korea

PhilippinesTaiwanThailandVenezuela

CambodiaIndonesia

Mozambique

Nepal

Nicaragua

CanadaChile

Costa Rica

Ecuador

FranceGuatemala

Iran

ItalyLiberia

Morocco

Netherlands

Spain

TunisiaTurkey

United Kingdom

Uruguay

Algeria

China

Dominican Republic

Jamaica

LaosMadagascar

Somalia

Sri Lanka

Trinidad and Tobago

Belgium

BulgariaCuba

Denmark

FinlandGreece

Malaysia

Montenegro

New Zealand

NorwayOman

Paraguay

Romania

Saudi Arabia

SerbiaSingapore

56

78

9Ln

GD

P pe

r cap

ita

0 .2 .4 .6 .8Female Political Empowerment

Mexico

SwedenSwitzerland

Ghana

South Africa

Japan

Burma/Myanmar

Russia

Albania

Egypt

Yemen

ColombiaPoland

Brazil

United States of America

Portugal

El Salvador

Bolivia

HaitiHonduras

Mali

PeruSenegal

SudanVietnamAfghanistan

Argentina

Ethiopia

IndiaKenya North KoreaSouth Korea

Lebanon

PhilippinesTaiwanThailandUganda

Venezuela

BeninBhutan

Burkina FasoCambodia

IndonesiaMozambique

Nepal

Nicaragua

NigerZambiaZimbabwe

Guinea

Ivory Coast

Mauritania

CanadaAustralia

BotswanaBurundiCape Verde

Central African Republic

Chile

Costa RicaEcuador

FranceGermany

GuatemalaIran Iraq

IrelandItaly

Jordan

Lesotho

Liberia

MalawiMongolia

Morocco

Netherlands

Panama

Qatar

Spain Syria

Tunisia

Turkey

United Kingdom

Uruguay

AlgeriaAngola

Chad ChinaDemocratic Republic of the Congo

Republic of the CongoDjibouti

Dominican Republic

Gabon

The Gambia

Guinea-Bissau

Laos

Madagascar

Namibia

Rwanda

SomaliaSri Lanka

SwazilandTogo

German Democratic RepublicAustria

Bahrain

Belgium

Bulgaria

Comoros

Cuba

Czech Republic

Denmark

Equatorial Guinea

Finland

Greece

Iceland

Israel

Kuwait

Luxembourg

Malaysia

Mauritius

New ZealandNorway

Oman

Paraguay Romania

Sao Tome and Principe

SerbiaSeychellesSingapore Hungary

67

89

10Ln

GD

P pe

r cap

ita

0 .2 .4 .6 .8Female Political Empowerment

Mexico

SwedenSwitzerland

Ghana

South Africa

Japan

Burma/Myanmar

Russia

AlbaniaEgypt

Yemen

Colombia

PolandBrazil

United States of America

Portugal

El Salvador

Bangladesh

Bolivia

Haiti

Honduras

Mali

Pakistan

Peru

SenegalSudan Vietnam

Afghanistan

Argentina

Ethiopia

IndiaKenyaNorth Korea

South Korea

KosovoLebanon

Nigeria

Philippines

Tanzania

Taiwan

Thailand

Uganda

Venezuela

Benin

Bhutan

Burkina FasoCambodia

Indonesia

MozambiqueNepal

Nicaragua

Niger

Zambia

Zimbabwe

Guinea

Ivory CoastMauritania

CanadaAustralia

Botswana

Burundi

Cape Verde

Central African Republic

ChileCosta Rica

Timor-Leste

Ecuador

FranceGermany

GuatemalaIran

Iraq

IrelandItaly

Jordan

Latvia

Lesotho

Liberia Malawi

Maldives

Mongolia

Morocco

Netherlands

Panama

Papua New Guinea

Sierra Leone

Spain

Syria

TunisiaTurkey

Ukraine

United Kingdom

Uruguay

Algeria

Angola

ArmeniaAzerbaijan

Belarus

Cameroon

Chad

China

Democratic Republic of the Congo

Republic of the CongoDjibouti

Dominican Republic

Eritrea

Gabon

The Gambia

Georgia

Guinea-Bissau

JamaicaKazakhstan

KyrgyzstanLaos

Libya

Madagascar

Moldova

Namibia

RwandaSomalia

Sri LankaSwaziland

TajikistanTogo

Trinidad and Tobago

TurkmenistanUzbekistan

Austria

Bahrain Barbados

Belgium

Bosnia and HerzegovinaBulgaria

Comoros

Croatia

Cuba

CyprusCzech Republic

Denmark

Equatorial GuineaEstonia

Fiji

FinlandGreece

Guyana

IcelandIsraelKuwait

Lithuania

Luxembourg

Macedonia

MalaysiaMauritius

New Zealand

Norway

ParaguayRomania

Sao Tome and Principe

Serbia

Seychelles

Singapore

SlovakiaSlovenia

Solomon Islands

Vanuatu

United Arab Emirates

Hungary

67

89

1011

Ln G

DP

per c

apita

0 .2 .4 .6 .8 1Female Political Empowerment

Figure 3: Scatter-plots, overlaid with (bivariate) best-fit lines and 95% confidence inter-vals, for Female Political Empowerment (data taken from V-Dem; x-axes) and Ln GDP percapita(data taken from Farris 2017; y-axes) in the years 1830 (top-left), 1900 (top-right),1950 (bottom-left) and 2000 (bottom-right).

average, between countries that have low and high female political empowerment. When

dividing the 21,853 observations into quartiles on the FPE index – with 0.172 marking the

cut-off for the lowest quartile, 0.344 the median, and 0.611 the highest quartile – we find

that average growth rates, based on the Fariss et al. data, increase monotonically and quite

substantially with FPE quartiles. The lowest quartile of FPE observations has an average

GDP per capita growth rate of 0.2 percent, and the second quartile grows, on average, at 0.6

percent. In contrast, the third quartile exhibited average growth of 1.5, whereas the upper

quartile grew at 2.7. When using the Maddison time series (Bolt and van Zanden, 2013),

which has numerous missing observations particularly among colonies and 19th century

countries, the corresponding growth rates are, respectively, 1.2, 1.2, 2.0, and 2.9. Countries

where women are politically empowered display much higher economic growth, on average.

20

Yet, the strong, positive correlation may stem from various sources, including the reverse

causal relationship and that different (observable or unobservable) confounders systemat-

ically affect both female empowerment and growth in particular directions. For instance,

political-historical and cultural characteristics that are prevalent in certain countries (e.g., in

Western Europe and North America) could enhance both female empowerment and growth.

Alternatively, confounding may come from time-specific factors; certain decades of modern

history may have given birth to ideological or technological trends that boosted female em-

powerment as well as growth. For these reasons, our benchmark OLS specification includes

both country- and year-fixed effects. Further, we cluster errors by country to account for

panel-level heteroskedasticiy and autocorrelation.

The theoretical discussion suggested that substantial time may pass before the hypothe-

sized effect from female empowerment is transmitted – via public policies and, in turn, their

impact on the behavior of firms and other economic agents – to technological change and

observed GDP per capita growth rates. While the exact lag-time of these processes are hard

to theorize and identify, we assume a five-year lag in our benchmark. We also test specifi-

cations where we measure growth closer (in time) to or further away from the independent

variables. We start out by analyzing country-years as unit of analysis to capture as much

information as possible and thus maximize efficiency. Yet, we also try out 5- and 10-year

panel structures, which have the benefits of smoothing out measurement errors and further

mitigating autocorrelation.

Concerning other covariates than the country- and year-fixed effects, we intentionally

keep our benchmark specification sparse to minimize missing due to listwise deletion and,

more importantly, mitigate post-treatment bias. The latter concern pertains to the possi-

bility that variables such as democracy or state capacity may be affected by female political

empowerment. Controlling for such institutional features could thus “block off” relevant in-

direct effects that we want to capture as part of our estimated, overall relationship. Hence,

the only additional covariate in our benchmark is initial Ln GDP per capita, as richer coun-

21

tries likely have better track-records of female empowerment, but also – due to standard

conditional convergence mechanisms (Barro and Sala-i Martin, 2004) – slower growth rates.

In alternative specifications, we introduce more covariates that may – even if we risk intro-

ducing post-treatment bias – also act as confounders. One example of such a (questionable)

extra control is regime type, as democracy may both causally affect female empowerment

and be influenced by it. (Indeed, even the conceptual boundaries between democracy and

FPR are unclear, as both consider, e.g., political participation and protection of rights.)

Hence, we test models excluding and models including democracy, as measured by V-Dem’s

Polyarchy index (Teorell et al., 2019)

Model 1.1 in Table 2 is the benchmark OLS specification on growth with country-year

as unit of analysis and using the GDP data from (Fariss et al., 2017). As discussed, this

model controls for country- and year- fixed effects in addition to initial Ln GDP per capita.

The dependent variable is the annual percentage change in GDP per capita five years after

independent variables are measured, i.e., the growth rate in t + 5. The model draws on

15,879 country-year observations from 182 countries, with maximum time series extending

across 221 years. As expected, there is a positive relationship between FPE and growth,

which is statistically significant at the 1% level. The point estimate indicates that going

from the first quartile score on FPE (.20; e.g., Italy under Mussolini in the 1930s) to the

third quartile score (.61; Australia in the 1950s) increases annual GDP per capita growth

with about 0.9 percentage points. The long-term consequences of such a difference in growth

rates are substantial. Consider two countries, A and B, that start out with identical GDP

per capita levels; and where A starts growing at a 0.9 percentage point higher rate. After

10 years, A’s GDP per capita is about 9 percentage points higher than B’s. After 20 years,

this difference has increased to 20 percent, and after 40 years to 43 percent. If we consider

an even larger change in FPE, going from the 10th percentile (.11; The Two Siciles in the

1820s or Sudan in the 1920s) to the 90th (.82; Canada or New Zealand in the 1970s), the

corresponding numbers for, respectively, 10, 20 and 40 years are differences in GDP per

22

capita of about 16, 35 and 84 percent. If the estimates of Model 1.1 are on point, improving

female political empowerment has substantial consequences for long-term developments in

income. As shown by the equivalent Model 1.2, which uses GDP data from Maddison, the

result is robust to using different data sources for measuring income level and growth.

Table 2: Main analysis: Fixed effects OLS regressions on GDP per capita growth or Ln GDPp.c. measured in t + 5

(1.1) (1.2) (1.3) (1.4) (1.5) (1.6)DV: GDP p.c. growth in year t+5 DV: Ln GDP p.c. in year t+5

Panel length 1 yr 1 yr 1 yr 1 yr 5 yrs 5yrsGDP data source Fariss Maddison Fariss Maddison Fariss Maddison

b/(t) b/(t) b/(t) b/(t) b/(t) b/(t)Female pol. empowerment 2.158*** 1.967** 0.110*** 0.114** 0.151*** 0.123**

(2.719) (2.151) (2.763) (2.508) (2.915) (2.439)Ln GDP per capita -1.237*** -2.533*** 0.937*** 0.892*** 0.922*** 0.888***

(-5.101) (-8.268) (64.898) (64.526) (45.493) (63.205)Country dummies Y Y Y Y Y YYear dummies Y Y Y Y Y YN 15879 13362 15880 13364 3165 2762Countries 182 163 182 163 180 162Max time series 221 215 221 215 44 44R2 0.029 0.085 0.945 0.948 0.933 0.947

Notes: ∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01. Errors are clustered by country. Covariates measured 5 years before DV.

One alternative way of specifying growth models, which is popular among growth economists

and has some distinct advantages (and disadvantages) from using GDP per capita growth

as dependent variable, is using forward-lagged Ln GDP per capita level as the dependent

variable.7 This alternative specification is less affected by measurement errors and other

sources of “noise” (business cycle fluctuations, big changes in prices for export products,

natural disasters, etc.), since GDP per capita growth rates can fluctuate widely from year

to year. Further, employing forward-lagged Ln GDP p.c. in, say, t + 5 instead of GDP

per capita growth from t + 4 to t + 5 allows us to also capture any short-term effects that

might exist, since we capture changes in income across the entire period from when initial

Ln GDP is measured (t) rather than only growth at the period’s end. The downside is that

this specification magnifies autocorrelation problems, since GDP levels display much higher

correlation with past GDP levels compared to current and past growth rates, and that it

7Mathematically, estimating the coefficient of a variable of interest on Ln GDP per capita level in t + 1,controlling for Ln GDP per capita in t, is equivalent to estimating the coefficient of this variable on GDPper capita growth from t to t + 1 (Hoeffler, 2002).

23

models conditional convergence dynamics less well – growth is no longer a linear function of

past income levels. Given these pros and cons of these different specifications, we opted to

test both versions. Models 1.3 and 1.4 thus replicate Models 1.1 and 1.2, respectively, but

with Ln GDP per capita in t+ 5 as the dependent variable. FPE remains robust also to this

specification choice.

Yet, we noted how using Ln GDP per capita as dependent variable magnifies autocor-

relation issues, which may influence results even if we cluster the errors by country. To

mitigate this issue, we followed another conventional approach used by growth economists

and re-estimated Models 1.3 and 1.4 on 5-year panels. When measuring the dependent vari-

able with 5-year intervals, there is weaker correlation with past realizations of the outcome

than when measuring it with 1-year intervals. Results are reported in Models 1.6 and 1.7,

and once again FPE is robust.

In sum, neither the source of GDP data, the exact specification of the dependent variable,

nor choice of panel structure affects the main result; there is a clear relationship between

female political empowerment and subsequent economic growth.

5.2 Robustness tests

In order to further investigate sensitivity, we ran additional robustness tests. Most of these

tests are reported in the Appendix, but we present a selection of important ones in Table 3.

Model 2.1 is equivalent to Model 1.1 from Table 2. This is our benchmark using Fariss et

al. GDP data and growth as dependent variable, and we include it in Table 3 simply to ease

comparisons with the alternative specifications.

First, we investigated whether the FPE result is sensitive to choice of lag-structure (see

also Figure 4). Model 2.2 alters the benchmark by measuring growth only one year after

the covariates. This specification is thus intended to capture only short-term effects. FPE

remains highly significant (t = 3.3), and the coefficient actually increases in size, suggesting

a strong boost in short-term growth from improved female empowerment. This finding is

24

somewhat surprising, given the theoretical discussion on the time it may take for changes in

political features to translate into new policies and subsequently technological change.

-4-2

02

46

t-10 t-5 t-3 t-1 t=0 t+1 t+3 t+5 t+10

Figure 4: Benchmark OLS Model (similar to 1.1, Table 2), but with GDP p.c. growthmeasured with various leads and lags relative to Female Political Empowerment Index. Whenlag structure is indicated as t− x [t + x], outcome is measured before [after] covariates.

Indeed, the latter result raises concerns that FPE might be correlated with trends in

growth due to other causal patterns (e.g., reverse causality) than the theorized effect. How-

ever, we tested whether FPE is correlated with contemporaneous growth, and it is not; FPE

is statistically insignificant and even flips sign (see Figure 4). Moreover, FPE is not related

to growth when the latter is measured before the former. We tested both 1-, 3-, 5- and

10-year lags on growth and neither are systematically correlated with current FPE; t-values

vary between -0.4 and +0.7 and coefficients are small in magnitude. We will return to al-

ternative specifications that deal with endogeneity concerns related to past trends in growth

confounding the relationship. In brief, we do not find evidence that such patterns are driving

25

our result. Hence, the most plausible interpretation of the result in Model 2.2 is, in fact,

that there exists a short-term effect of FPE on growth, in addition to the (theoretically less

surprising) medium-term effect captured by our benchmark.

As Models 2.3 (growth measured in t+ 3) and 2.4 (growth measured in t+ 10) show, the

relationship is also robust to assuming alternative intermediate and longer-term effect lags.

The FPE coefficient remains statistically significant at 5% and sizeable, even if somewhat

attenuated relative to the benchmark.

As discussed, we kept our benchmark sparse to mitigate post-treatment bias. We did not,

for example, control for democracy level, as this might be affected by female empowerment

– controlling for democracy could thus block a relevant indirect effect. However, more

democratic regimes might more easily experience improvements in female empowerment,

for example because female voters are susceptible to vote to reward, or otherwise pressure,

parties into adopting political and institutional changes that enhance female representation

in politics. If democracy also enhances growth (Gerring et al., 2005; Acemoglu et al., 2019),

omitting democracy from the regression could lead to (upward) omitted variable bias for

FPE. Model 2.5 therefore adds V-Dem’s Polyarchy index to the benchmark. Holding scores

on this index constant has the additional benefit of accounting for potential subjective coder

biases in the V-Dem data. For example, fast-growing economies could be evaluated in

an artificially positive manner on different political measures by country-experts, if strong

economic performance induces overall positive impressions of the country’s politics. Such

a bias would, presumably, jointly affect FPE and Polyarchy, and controlling for Polyarchy

should thus purge the FPE coefficient of this bias. Yet, FPE remains significant at 5%

and actually increases somewhat in size, from 2.2 to 2.6, when controlling for Polyarchy.

Likewise, FPE retains a value of 2.2 and is significant at 5% in Model 2.6, which controls for

a proxy of state capacity, namely V-Dem’s indicator on impartial and rule-following public

administration.

26

Tab

le3:

Sel

ecte

dro

bust

nes

ste

sts

(2.1)

(2.2)

(2.3)

(2.4)

(2.5)

(2.6)

(2.7)

(2.8)

(2.9)

(2.10)

(2.11)

(2.12)

(2.13)

(2.14)

1-yea

rpanels;

DV:GDP

p.c.growth

5-yea

rpanels;

DV:LnGDP

p.c.

DV

mea

suredin

t+

5t+

1t+

3t+

10

t+

5t+

5t+

5t+

5t+

5t+

5t+

10

t+

5t+

5t+

5b/(t)

b/(t)

b/(t)

b/(t)

b/(t)

b/(t)

b/(t)

b/(t)

b/(t)

b/(t)

b/(t)

b/(t)

FPE

2.158***

3.785***

1.810**

1.968**

2.629**

2.172**

1.453

2.141***

2.495*

0.151***

0.278***

0.166***

0.329**

0.314**

(2.719)

(3.335)

(2.306)

(2.025)

(2.567)

(2.434)

(1.351)

(2.715)

(1.937)

(2.915)

(3.117)

(3.293)

(2.151)

(2.099)

LnGDP

p.c.

-1.237***

-2.402***

-0.967***

-1.484***

-1.291***

-1.239***

-1.637***

-1.197***

-1.798***

0.922***

0.852***

1.037***

0.989***

1.030***

(-5.101)

(-4.044)

(-4.328)

(-5.074)

(-4.998)

(-4.909)

(-6.058)

(-5.134)

(-5.696)

(45.493)

(25.255)

(10.618)

(50.431)

(6.467)

Polyarchy

-0.579

-0.824

(-0.817)

(-0.943)

Imp.publicadm.

-0.001

-0.063

(-0.012)

(-0.438)

Resourcedep

.-0.038**

-0.038**

(-2.209)

(-2.178)

Lnpopulation

0.241

-0.306

(1.041)

(-1.096)

LnGDP

p.c.t−

5-0.105

-0.039

(-1.102)

(-0.246)

LnGDP

p.c.t−

10

-0.007

(-0.268)

LnGDP

p.c.t−

15

0.003

(0.154)

Countrydummies

YY

YY

YY

YY

YY

YY

Yea

rdummies

YY

YY

YY

YY

YY

YY

YY

N15879

16645

16256

14947

15552

15857

10716

15879

10510

3165

2987

2732

3165

3018

R2

0.029

0.035

0.027

0.031

0.029

0.028

0.135

0.029

0.139

0.933

0.877

0.949

Instru

men

ts126

126

HansenJ-test

.76

.81

Ar(2)

.08

.86

AR(3)

.63

.48

Notes:

∗p<0.1;∗∗

p<0.05;∗∗

∗p<0.01.Errors

are

clustered

byco

untry.

GDP

data

are

from

Fariss

etal.

(2017).

Models1–12are

estimatedbyOLSandModels13-14bySystem

GMM.

27

Model 2.7 adds a measure of natural resource dependence (share of GDP from oil, natural

gas, coal and minerals) from Miller (2015) to the benchmark. While FPE remains sizeable,

with a coefficient of 1.5, it is no longer statistically significant at conventional levels (t = 1.4).

However, further analysis shows that much of this attenuation is due to the changing sample;

missing data on the resource dependence variable truncates the sample from 15,879 to 10,716

observations. When re-running the benchmark (Model 2.1) on this reduced sample, the FPE

point-estimate is 1.8 and the t-value is 1.6. In Model 2.8, which controls for Ln population

and where the sample is again expanded to 15,879 observations, FPE is similar in size

and significance to the benchmark. Moreover, in the “kitchen-sink” specification (Model

2.9), which simultaneously controls for Polyarchy, impartial administration, natural resource

dependence and population, FPE is actually higher than in the benchmark (2.5) with a

p-value of 0.055.

We conducted similar tests for 5-year panel specifications using Ln GDP per capita as

dependent variable. Model 2.10 replicates Model 1.5 from Table 2, measuring the dependent

variable five years after the covariates. Model 2.11 maintains the 5-year panel set-up, but

measures the outcome 10 years (i.e., two panel periods) after the covariates. This change

actually strengthes the relationship quite substantially, increasing the FPE coefficient from

0.15 to 0.28. This change was to be expected with Ln GDP p.c. as dependent variable.

With income being measured at the end of a 10-year period, we capture both the shorter-

and medium-term effects of a change in FPE. Further, in Model 2.12, we tested for the

potential endogeneity to prior trends in income growth by including three additional lags

of the dependent variable (t − 5, t − 10, and t − 15, in addition to Ln GDP p.c. measured

in t; the dependent variable is measured in t + 5). By doing so, we follow the approach by

Acemoglu et al. (2019) to account for pre-treatment patterns in income growth. The FPE

coefficient and corresponding t-value increase somewhat when doing so. Once again, there

are few empirical indications that historical trends in income growth confound the observed

relationship between FPE and subsequently measured growth.

28

Still, we conducted one additional type of test to account for potential endogeneity in

FPE, running so-called System Generalized Method of Moments (GMM) models. These

models are attuned to estimating relationships involving sluggish variables such as FPE (see

Blundell and Bond, 1998). In System GMM specifications, lags of differences in variables are

used to instrument for current levels, and, likewise, lags of levels are used to instrument for

current differences. The specifications reported in Table 3 model only FPE as endogenous

and use the second and third lags for instrumentation (to keep the instrument count below

the the number of cross-section units, see Roodman, 2009).8 Alternative specifications are

reported in Appendix Table A.3. Model 2.13 includes only the first lag of the dependent

variable as regressor, whereas 2.14 includes the first and second lags. Both specifications

report a statistically significant FPE coefficient that is substantially higher than in the OLS

models. While the Ar(2)-test in Model 2.13 gives some reason for concern that residual

autocorrelation affects results, the specification tests for Model 2.14 suggest that this model

may yield a consistent estimate of the effect of FPE on growth. This estimate is both positive

and substantially sizeable.

In sum, there is fairly strong evidence, from different panel regressions, that improvements

to female political empowerment enhances subsequent economic growth.

5.3 Assessing potential heterogeneity in the relationship

We wanted to check if the identified pattern is consistent across time periods, geographical

contexts, and political regime types, or if it is much stronger in some settings than oth-

ers. For these particular tests on heterogeneity, our theoretical argument does give clear

a priori expectations. If anything, we expected the relationship to be relatively persistent

across contexts, as our argument points to mechanisms that should not be contingent on

being in a particular region, historical period or regime type – female political empowerment

should enhance the variation in new ideas and the selection of more efficient ones in different

8The instrument count is also a function of number of panel units. Hence, well-specified GMM modelsare impossible to achieve for the country-year set-up.

29

contexts.

We ran different tests to assess potential heterogeneity in the relationship (see Appendix

Table A.4 for interaction model tests). Figure 5 (top panel) presents regression coefficients

on FPE with 95% confidence intervals from straightforward split-sample tests, where we

estimate the benchmark Model 1.1., but on limited samples. The leftmost coefficients per-

tain to, respectively, pre- and post-WWII samples. The two middle coefficients pertain to,

respectively, “Western” (Western European countries, Canada, United States, Australia,

and New Zealand) and all other countries. Finally, the two rightmost coefficients pertain

to, respectively, democracies and autocracies, where we use the Lexical Index of Electoral

Democracy by Skaaning, Gerring and Bartusevicius (2015) and require competitive elections

and universal suffrage for coding a regime as democratic. Interestingly, results indicate a

somewhat stronger and clearer estimated effect of female empowerment in autocratic con-

texts (where the rate of technological change is, overall, lower Knutsen, 2015). Further,

the estimated relationship is stronger for the pre-WWII era than the period from 1946 and

onwards, and the relationship is present and clear in “non-Western” countries, but not in

“Western” countries.

We note, however, that the split sample results based on fewer units and/or shorter time

series are sensitive to specification choices. For instance, when omitting country-fixed effects,

the relationship is large and highly significant also across the 31 included Western countries

(as well as across the 151 non-Western ones). And, when omitting year-fixed effects the

relationship is large and significant both in the pre-WWII and post-WWII samples. The

bottom panel of Figure 5 shows fairly similar results when omitting both the country-and

year-fixed effects simultaneously. In this specification, the estimated relationship is also

virtually similar (and highly significant) across regime types. While we would not put too

much trust in the latter estimates, with omitted variables biases possibly affecting results,

they serve to illustrate the more sensitive nature of the split-sample results.

30

-20

24

6

1789-1945 1946-2010 Western Non-Western Autocratic Democratic

24

60

-2

1789-1945 1946-2010 Western Non-Western Autocratic Democratic

Figure 5: Coefficient plots with 95% confidence intervals for Female Political EmpowermentIndex on limited samples, restricted by time period, geography, or regime type. Top panel:Coefficients are from equivalents to benchmark Model 1.1, Table 2. Bottom panel: Cor-responding coefficients for specifications that omit country- and year-fixed effects, but areotherwise similar to Model 1.1., Table 2.

Next, we assessed whether the finding for the composite FPE index is driven by one

particular sub-component. We remind that the aggregated index consists of three sub-

indices – on civil liberties, political participation, and civil society participation – capturing

distinct aspects of female empowerment, and that our theoretical argument suggested that

31

all three aspects should contribute to enhance technological change, and thus growth.

01

23

4

Female Pol. Empow. Index Civil liberties index Civil society participation index Political participation index

Figure 6: Coefficient plots with 95% confidence intervals for Female Political EmpowermentIndex and its three sub-indices. All coefficients are from equivalents to the benchmark Model1.1, Table 2, with annual GDP per capita growth in t+ 5 as dependent variable. All indicesrange from 0–1.

As anticipated, all three sub-indices are individually related to subsequent growth rates.

Figure A.1 displays coefficient plots with 95% confidence intervals for specifications akin to

the benchmark Model 1.1, Table 2. As for the aggregated FPE, all indices range from 0 to 1,

and coefficients are thus comparable. Interestingly, the composite index estimate is around

twice the size of the sub-indices, which are strikingly similar in size. Going from minimum

to maximum on any of these sub-indices is predicted to increase GDP per capita growth

rates by 1.0 to 1.2 percentage points, and all three coefficients are at least weakly significant

(10% level).9 Moreover, all three sub-indices, and especially civil liberties and political

9When we include all three sub-indices simultaneously in the benchmark, the civil society index turnsvery close to 0, whereas the civil liberties (1.2; t = 1.7) and political participation (0.9; t = 1.4) indicesbasically retain their sizes but obtain lower t-values.

32

participation, are quite robust to specification changes such as altering the dependent variable

specification (Ln GDP p.c. instead of GDP p.c. growth) or using Maddison GDP data. In

sum, there is evidence that the different, theorized aspects of female empowerment carry an

independent relationship with subsequent growth, and that the relationship is even stronger

for the composite concept than for any of the individual sub-components.

5.4 Female empowerment and technological change

Finally, we turn to investigating another, and more specific, implication from our argument,

namely that improvements in female political empowerment should enhance technological

change. That is, even when disregarding economic growth coming from investments in

physical or human capital, we should find that FPE is related to higher residual growth, which

is presumably driven mainly by improvements in production or organization technologies.

To test this expectation, we draw on the above-described, extensive TFP data from

Baier, Dwyer Jr. and Tamura (2006). To recapitulate, these data were arrived at via

a comprehensive growth accounting exercise on 145 countries and the longest time series

(United Kingdom) extends from 1831–2000. When following the interpolation procedure

used by Knutsen (2015), we can re-run our benchmark – substituting Ln GDP per capita

from Model 1.3, Table 2 with Ln TFP – on 6827 country-year observations from 142 countries.

This specification is reported as Model 3.1 in Table 4. It reveals a positive and significant

(5% level) relationship between FPE and Ln TFP measured five years later, conditional on

initial level of Ln TFP and country- and year-fixed effects. As for Ln GDP per capita, the

FPE coefficient increases markedly in size (from .10 to .16) once measuring the outcome 10

years instead of 5 years after the covariates in Model 3.2. This latter observation suggests

that there are longer term benefits to technological efficiency from increased female political

empowerment, beyond the notable short-term benefits.

Yet, we discussed for the growth regressions how certain variables, notably including

democracy level, may be relevant controls, even if adding them could introduce post-treatment

33

Table 4: Fixed effects OLS regressions on Ln Total Factor Productivity (TFP)

(3.1) (3.2) (3.3) (3.4) (3.5) (3.6) (3.7)1-year panels 5-year panels

DV measured in t+ 5 t+ 10 t+ 5 t+ 5 t+ 5 t+ 5 t+ 5b/(t) b/(t) b/(t) b/(t) b/(t) b/(t) b/(t)

Female political empowerment 0.097** 0.157** 0.077 0.077* 0.087** 0.108** 0.109*(2.274) (2.023) (1.387) (1.693) (2.059) (2.031) (1.657)

Ln TFP 0.931*** 0.636*** 0.933*** 0.941*** 0.909*** 0.938*** 0.940***(37.354) (12.490) (35.683) (31.914) (32.693) (36.227) (34.915)

Polyarchy 0.018 -0.001(0.468) (-0.022)

Resource dependence -0.002***(-2.743)

Ln population -0.056**(-2.140)

Country dummies Y Y Y Y Y Y YYear dummies Y Y Y Y Y Y YN 6827 6124 6790 6227 6689 1456 1447R2 0.838 0.606 0.837 0.839 0.832 0.825 0.824

Notes: ∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01. Errors are clustered by country.

bias for FPE. The same points apply here. Model 3.3 therefore includes V-Dem’s Polyarchy

Index as an additional control, and this attenuates the FPE coefficient by about 20 percent

(from 0.097 to 0.077) relative to the benchmark. Moreover, FPE now turns statistically

insignificant at conventional levels (t = 1.4). We also tested a model controlling for V-Dem’s

impartial public administration described above, and – similarly to the results for Model

3.3 controlling for Polyarchy – the relationship turns insignificant. In contrast, the positive

relationship between FPE and Ln TFP measured five years later is robust to controlling for

natural resource dependence (Model 3.4) or population size (Model 3.5). Still, and in con-

trast with the GDP per capita regressions, the FPE coefficient on TFP growth is sensitive

to the control strategy.

The lack of robustness for the relationship with TFP growth may stem from different

factors. First, the GDP regressions included far more country-year observations, thus giving

us more efficient estimates and a lower likelihood of conducting Type II errors. Further,

the results for Models 3.1–3.5 may be weakened by measurement error induced from the

interpolation routine in the country-year panels. The TFP data are originally measured

with intervals of several years (typically 10) between each data point, and the interpolation

procedure may artificially smooth out growth across these longer intervals. Hence, Models

34

3.6 (benchmark) and 3.7 (adding Polyarchy) employ 5-year panel units. While results are

fairly similar, the FPE coefficient is somewhat larger in size and now weakly significant in

the model including Polyarchy. In sum, there is some, but not completely robust, evidence

suggesting that FPE is related to faster subsequent technological change, as measured by

TFP growth.

6 Conclusion

We have argued that political institutions that enhance key aspects of female political em-

powerment – pertaining to the representation, voice, and active participation of women in

politics and civil society – influence a country’s rate of technological change. Such em-

powerment should enhance technological change both through affecting the variety of new

ideas introduced into the economy as well as the selection of the more efficient ideas. Since

technological change is the key “immediate determinant” of long-term economic growth,

we also anticipate that female political empowerment has a substantial effect on economic

development.

Drawing on data from 182 countries and time series extending back to 1789, we found

fairly robust evidence for different implications from our argument. The most robust evi-

dence pertains to the implication that female political empowerment is positively related to

subsequent economic growth. This relationship holds up for different measurement strategies

and statistical specifications. Reassuringly, the relationship holds up when accounting for

country- and year-fixed effects, and female political empowerment is only correlated with

subsequently measured growth and not with contemporaneous or previous growth rates.

Second, as anticipated from our argument, measures capturing all three sub-components of

female political empowerment are individually linked to growth. Third, we also find rela-

tively (though not completely) robust evidence indicating that female political empowerment

relates to TFP growth, an indicator of technological change.

35

Our results could have real-world relevance, insofar as some decision makers are more

concerned with economic performance than questions of justice and equity in representation,

inclusion, and protection of civil liberties across. We show that the inclusion of women in

politics may not only be justified by intrinsic, normative motives. There is also a more

instrumental “business case” for female empowerment, which might sway otherwise reluctant

social groups and decision makers to work for, or at least acquiesce to, the inclusion of women

in decision-making.

There is persistent under-representation of women in politics and lacking protection of

women’s civil liberties. Despite some progress, worldwide, in recent decades, countries – and

especially middle and low income countries – still vary a lot in female political empowerment

(World Bank, 2020b). In many places, there are still substantial restrictions on women’s

opportunities to participate in the economy on equal footing with men. Around a third of

the world’s countries restrict the freedom of movement for women, 40 percent have legal

restrictions on women’s decisions to join and remain in the labor force, and around 40

percent discriminate against women in their property rights legislation. Finally, in 115

countries, World Bank experts judge, women cannot run a business in the same way as

men (World Bank, 2020b). These striking inequalities are reflected in the composition of

the political institutions that have the power to influence the relevant legislation. In 2019,

only 24.6 on parliamentary seats worldwide were held by women (World Bank, 2020a), and

the corresponding number for cabinet seats was 20.7 (Union and women, 2019). Still today,

there is ample room for increased representation of and participation by women in many

countries. Our results suggest that, in these countries, there is also corresponding room for

more rapid technological change and economic development.

36

References

Acemoglu, D. 2008. Introduction to Modern Economic Growth. Princeton: Princeton Uni-

versity Press.

Acemoglu, Daron and James A. Robinson. 2012. Why Nations Fail: The Origins of Power,

Prosperity and Poverty. New York: Crown Business.

Acemoglu, Daron, Simon Johnson and James A. Robinson. 2001. “The Colonial Origins

of Comparative Development: An Empirical Investigation.” American Economic Review

91(5):1369–1401.

Acemoglu, Daron, Suresh Naidu, Pascual Restrepo and James A. Robinson. 2019. “Democ-

racy does cause growth.” Journal of Political Economy 127(1):47–100.

Aghion, Phillipe and Peter Howitt. 1992. “A Model of Growth through Creative Destruc-

tion.” Econometrica 60:323–351.

Baier, Scott L., Gerald P. Dwyer Jr. and Robert Tamura. 2002. “How Important Are Capital

and Total Factor Productivity for Economi Growth?” Federal Reserve Bank of Atlanta.

Working Paper 2002-2a.

Baier, Scott L., Gerald P. Dwyer Jr. and Robert Tamura. 2006. “How Important Are Capital

and Total Factor Productivity for Economi Growth?” Economic Inquiry 44(1):23–49.

Barro, Robert J. and Xavier Sala-i Martin. 2004. Economic Growth. Cambridge, MA.: The

MIT Press.

Beaman, Lori, Raghabendra Chattopadhyay, Esther Duflo, Rohini Pande and Petia

Topalova. 2009. “Powerful Women: Does Exposure Reduce Bias?” Quarterly Journal

of Economics 124(4):1497–1540.

Birnir, Johanna Kristın and David M Waguespack. 2011. “Ethnic inclusion and economic

growth.” Party Politics 17(2):243–260.

37

Bizzarro, Fernando, John Gerring, Allen Hicken, Carl Henrik Knutsen, Michael Bernhard,

Svend-Erik Skaaning, Michael Coppedge and Staffan I Lindberg. 2018. “Party Strength

and Economic Growth.” World Politics 70(2):275–320.

Blundell, Richard and Stephen Bond. 1998. “Initial Conditions and Moment Restrictions in

Dynamic Panel Data Models.” Journal of Econometrics 87(1):115–143.

Bolt, Jutta and Jan L. van Zanden. 2013. “The First Update of the Maddison Project:

Re-Estimating Growth Before 1820.” Maddison Project Working Paper 4.

Bunch, Charlotte. 1990. “Women’s rights as human rights: Toward a re-vision of human

rights.” Human Rights Quarterly 12(4):486–498.

Chattopadhyay, Raghabendra and Esther Duflo. 2004. “Women as policy makers: Evidence

from a randomized policy experiment in India.” Econometrica 72(5):1409–1443.

Coppedge, Michael, John Gerring, Adam Glynn, Carl Henrik Knutsen, Stafan I. Lindberg,

Daniel Pemstein, Brigitte Seim, Svend-Erik Skaaning, Jan Teorell and et al. 2020. Varieties

of Democracy: Measuring Two Centuries of Political Change. Cambrdige: Cambridge

University Press.

Coppedge, Michael, John Gerring, Carl Henrik Knutsen, Staffan I. Lindberg, Jan Teorell,

David Altman, Michael Bernhard, Steven M. Fish, Adam Glynn, Allen Hicken, Anna

Luhrmann, Kyle L. Marquardt, Kelly McMann, Daniel Pemstein, Brigitte Seim, Rachel

Sigman, Svend-Erik Skaaning, Jeffrey Staton, Steven Wilson, Agnes Cornell, Lisa Gastaldi,

Haakon Gjerlow, Nina Ilchenko, Joshua Krusell, Laura Maxwell, Valeriya Mechkova, Juraj

Medzihorsky, Josefine Pernes, Johannes von Romer, Natalia Stepanova, Aksel Sundstrom,

Eitan Tzelgov, Yi-ting Wang, Tore Wig and Daniel Ziblatt. 2019. “V-Dem Country-Year

Dataset v.9.” Varieties of Democracy (V-Dem) Project.

Coppedge, Michael, John Gerring, Carl Henrik Knutsen, Staffan I. Lindberg, Jan Teorell,

David Altman, Michael Bernhard, Steven M. Fish, Adam Glynn, Allen Hicken, Anna

38

Luhrmann, Kyle L. Marquardt, Kelly McMann, Pamela Paxton, Daniel Pemstein, Brigitte

Seim, Rachel Sigman, Svend-Erik Skaaning, Jeffrey Staton, Agnes Cornell, Lisa Gastaldi,

Haakon Gjerlow, Valeriya Mechkova, Johannes von Romer, Aksel Sundstrom, Eitan Tzel-

gov, Yi-ting Wang, Tore Wig and Daniel Ziblatt. 2019. “V-Dem Codebook, v.9.” Varieties

of Democracy (V-Dem) Project.

Cuberes, David and Marc Teignier. 2012. “Gender gaps in the labor market and aggregate

productivity.” University of Sheffield: Working Paper.

Cuberes, David and Marc Teignier. 2014. “Gender inequality and economic growth: A

critical review.” Journal of International Development 26(2):260–276.

Dahlum, S., Carl-Henrik Knutsen and Staffan I. Lindberg. 2018. “Economic consequences

of civil liberties.” B-Team Research Paper.

De Long, J Bradford and Andrei Shleifer. 1993. “Princes and Merchants: European City

Growth Before the Industrial Revolution.” Journal of Law and Economics 36:671–702.

Diamond, Jared M. 1997. Guns, Germs and Steel: The Fates of Human Societies. New

York: W.W. Norton.

Duflo, Esther. 2012. “Women empowerment and economic development.” Journal of Eco-

nomic Literature 50(4):1051–79.

Esteve-Volart, Berta. 2004. “Gender discrimination and growth: Theory and evidence from

India.” London: SICERD Development Economics Discussion Paper 42.

Estrin, Saul and Tomasz Mickiewicz. 2011. “Institutions and female entrepreneurship.” Small

business economics 37(4):397.

Evans, Heather K and Jennifer Hayes Clark. 2016. ““You Tweet Like a Girl!” How Female

Candidates Campaign on Twitter.” American Politics Research 44(2):326–352.

39

Evans, Peter B. 1995. Embedded Autonomy. States and Industrial Transformation. Princeton:

Princeton University Press.

Evans, Peter and James E. Rauch. 1999. “Bureaucracy and Growth: A Cross-National

Analysis of the Effects of “Weberian” State.” American Sociological Review 64(5):748–

765.

Fariss, Christopher J., Charles D. Crabtree, Therese Anders, Zachary M. Jones, Fridolin J.

Linder and Jonathan N. Markowitz. 2017. “Latent Estimation of GDP, GDP per capita,

and Population from Historic and Contemporary Sources.” Working Paper.

Fukuyama, Francis. 2014. Political Order and Political Decay. London: Profile Books.

Gerring, John, Philip Bond, William T Barndt and Carola Moreno. 2005. “Democracy and

economic growth.” World Politics 57(3):323–364.

Ghani, Ejaz, Anandi Mani and Stephen D O’Connell. 2013. “Can political empowerment

help economic empowerment? Women leaders and female labor force participation in

India.” Policy Research Working Paper WPS 6675. Washington D.C.: World Bank.

Goldstein, Markus and Christopher Udry. 2008. “The profits of power: Land rights and

agricultural investment in Ghana.” Journal of political Economy 116(6):981–1022.

Goltz, Sonia, Mari W Buche and Saurav Pathak. 2015. “Political empowerment, rule of

law, and women’s entry into entrepreneurship.” Journal of Small Business Management

53(3):605–626.

Grossman, Gene M. and Elhanan Helpman. 1991. “Quality Ladders in the Theory of

Growth.” The Review of Economic Studies 58(1):43–61.

Helpman, Elhanan. 2004. The Mystery of Economic Growth. Cambridge, Ma.: The Belknapp

Press of Harvard University Press.

40

Hoeffler, Anke. 2002. “The augmented Solow model and the African growth debate.” Oxford

Bulletin of Economics and Statistics 64(2):135–158.

Kabeer, Naila and Luisa Natali. 2013. “Gender Equality and Economic Growth: Is there a

Win-Win?” IDS Working Papers 2013(417):1–58.

Kahn, Kim Fridkin. 1993. “Gender differences in campaign messages: The political ad-

vertisements of men and women candidates for US Senate.” Political Research Quarterly

46(3):481–502.

Kanter, Rosabeth Moss. 2008. Men and women of the corporation. New York: Basic books.

Khan, Sarah. 2017. “What Women Want: Gender Gaps in Political Preferences.” Compar-

ative Politics Newsletter 65(1):42.

Klasen, Stephan. 2002. “Low schooling for girls, slower growth for all? Cross-country ev-

idence on the effect of gender inequality in education on economic development.” The

World Bank Economic Review 16(3):345–373.

Klasen, Stephan and Francesca Lamanna. 2009. “The impact of gender inequality in ed-

ucation and employment on economic growth: New evidence for a panel of countries.”

Feminist Economics 15(3):91–132.

Klenow, Peter J. and Andres Rodriguez-Clare. 1997. The Neo-Classical Revival in Growth

Economics: Has it Gone to Far? In NBER Macroeconomics Annual 1997, ed. Ben S.

Bernanke and Julio J. Rotemberg. Cambridge, MA: MIT Press.

Knowles, Stephen, Paula K Lorgelly and P Dorian Owen. 2002. “Are educational gender

gaps a brake on economic development? Some cross-country empirical evidence.” Oxford

economic papers 54(1):118–149.

Knutsen, Carl Henrik. 2015. “Why democracies outgrow autocracies in the long run: Civil

liberties, information flows and technological change.” Kyklos (68):357–384.

41

Kremer, Michael. 1993a. “The O-Ring Theory of Economic Development.” The Quarterly

Journal of Economics 108(3):551–575.

Kremer, Michael. 1993b. “Population Growth and Technological Change: One Million B.C.

to 1990.” Quarterly Journal of Economics 108(3):681–716.

Landes, David S. 1998. The Wealth and Poverty of Nations. London: Abacus.

Longwe, Sara Hlupekile. 2000. “Towards realistic strategies for women’s political empower-

ment in Africa.” Gender & Development 8(3):24–30.

Mankiw, Gregory N., David Romer and David N. Weil. 1992. “A Contribution to the

Empirics of Economic Growth.” The Quarterly Journal of Economics 107(2):407–437.

Mansbridge, Jane. 1999. “Should blacks represent blacks and women represent women? A

contingent “yes”.” Journal of Politics 61(3):628–657.

Mechkova, V. and Ruth Carlitz. 2018. “Gendered Accountability: When and Why Do

Women’s Policy Priorities Get Implemented?” V-Dem Working Paper 88.

Mechkova, Valeriya and Steven Wilson. 2019. “Does Gender Still Matter for Politics? The

Case of the 2018 U.S. Elections on Twitter.” Digital Society Project Working Paper.

Miller, Grant. 2008. “Women’s suffrage, political responsiveness, and child survival in Amer-

ican history.” Quarterly Journal of Economics 123(3):1287–1327.

Miller, Michael K. 2015. “Democratic Pieces: Autocratic Elections and Democratic Devel-

opment Since 1815.” British Journal of Political Science 45(03):501–530.

Mitra, Aniruddha, James T Bang and Arnab Biswas. 2015. “Gender equality and economic

growth: Is it equality of opportunity or equality of outcomes?” Feminist Economics

21(1):110–135.

42

Nelson, Richard R. 2005. Technology, Institutions and Economic Growth. Cambridge, MA:

Harvard University Press.

Nelson, Richard and Sidney Winter. 1982. An Evolutionary Theory of Economic Growth.

Cambridge, MA: Harvard University Press.

North, Douglass C. 1990. Institutions, Institutional Change and Economic Performance.

Cambridge: Cambridge University Press.

North, Douglass C. 2005. Understanding the Process of Economic Change. Princeton: Prince-

ton University Press.

North, Douglass C., John Joseph Wallis and Barry R. Weingast. 2009. Violence and Social

Orders: A Conceptual Framework for Interpreting Recorded Human History. Cambridge:

Cambridge University Press.

Olson, Mancur. 1993. “Democracy, Dictatorship and Development.” American Political

Science Review 87(3):567–576.

Pemstein, Daniel, Kyle L Marquardt, Eitan Tzelgov, Yi-ting Wang, Joshua Krusell and

Farhad Miri. 2018. “The V-Dem measurement model: Latent variable analysis for cross-

national and cross-temporal expert-coded data.” V-Dem Working Paper 2018:21.

Phillips, Anne. 1995. The politics of presence. Oxford: Clarendon Press.

Pitkin, Hanna F. 1967. The concept of representation. Berkeley: University of California

Press.

Przeworski, Adam, Michael E. Alvarez, Jose Antonio Cheibub and Fernando Limongi. 2000.

Democracy and Development. Political Institutions and Well-Being in the World, 1950–

1990. Cambridge: Cambridge University Press.

43

Rodrik, Dani, Arvind Subramanian and Francesco Trebbi. 2004. “Institutions Rule: The Pri-

macy of Institutions Over Geography and Integration in Economic Development.” Journal

of Economic Growth 9(2):131–165.

Romer, Paul. 1990. “Endogenous Technological Change.” Journal of Political Economy

98(5):71–102.

Romer, Paul. 1993. “Idea Gaps and Object Gaps in Economic Development.” Journal of

Monetary Economics 32(3):543–573.

Roodman, David. 2009. “A Note on the Theme of Too Many Instruments.” Oxford Bulletin

of Economics and Statistics 71(1):135–158.

Rothstein, Bo and Jan Teorell. 2008. “What Is Quality of Government? A Theory of

Impartial Government Institutions.” Governance 21(2):165–190.

Skaaning, Svend-Erik, John Gerring and Henrikas Bartusevicius. 2015. “A Lexical Index of

Electoral Democracy.” Comparative Political Studies 48(12):1491–1525.

Solow, Robert M. 1956. “A Contribution to the Theory of Economic Growth.” The Quarterly

Journal of Economics 70:65–94.

Sundstrom, Aksel, Pamela Paxton, Yi-Ting Wang and Staffan I Lindberg. 2017. “Women’s

political empowerment: A new global index, 1900–2012.” World Development 94:321–335.

Swiss, Liam, Kathleen M Fallon and Giovani Burgos. 2012. “Does critical mass matter?

Women’s political representation and child health in developing countries.” Social Forces

91(2):531–558.

Teorell, Jan, Michael Coppedge, Staffan I. Lindberg and Svend-Erik Skaaning. 2019. “Mea-

suring Polyarchy Across the Globe, 1900–2017.” Studies in Comparative International

Development 54(1):71–95.

44

Thevenon, Olivier, Nabil Ali, Willem Adema and Angelica Salvi del Pero. 2012. “Effects of

reducing gender gaps in education and labour force participation on economic growth in

the OECD.” OECD Social, Employment and Migration Working Paper 138.

Union, Inter-Parliamentary and UN women. 2019. “Women in Politics: 2019.” Poster.

URL: http://www.unwomen.org/en/digital-library/publications/2019/03/women-in-

politics-2019-map

Verspagen, Bart. 2005. Innovation and Economic Growth. Vol. The Oxfors Handbook of

Innovation Oxford: Oxford University Press chapter 18, pp. 487–513.

World Bank. 2020a. “DataBank.” Online: databank.worldbank.org.

World Bank. 2020b. “Women, Business and the Law 2020.” Dataset.

Young, Alwyn. 1995. “The Tyranny of Numbers: Confronting the Statistical Realities of the

East Asian Growth Experience.” The Quarterly Journal of Economics 110(3):641–680.

Young, Iris Marion. 2011. Justice and the Politics of Difference. Princeton: Princeton

University Press.

45

A Online Appendix

In this appendix we present descriptive statistics and various robustness tests that were

mentioned, but not reported in tables or figures, in the paper. First, we present descriptive

statistics for all core variables used in the analysis, with histograms displaying the distri-

bution of our main independent and dependent variables. Second, we present additional

robustness tests employing the GDP data from Maddison. Thereafter, we present additional

GMM specifications on 5- and 10-year panels. Next, we show alternative tests on heterogene-

ity of the relationship across time, space, and regime type, using interaction models rather

than split-sample tests. Finally, we show robustness tests for the regressions with the three

sub-indices of our main Female Empowerment Measure as alternative independent variables.

i

Table A.1: Descriptive statistics for main variables used in our analysis

Variable (source) N Mean Std. dev.GDP p.c. growth (Fariss et al.) 20,622 1.202878 9.975378GDP p.c. growth (Maddison) 16,525 1.945244 7.374358Ln GDP p.c. (Fariss et al.) 20,828 7.602531 1.062199Ln GDP p.c. (Maddison) 16,695 8.048396 1.139924Ln TFP (Baier et al.) 8,462 4.842124 .4978569Female Political Empowerment (V-Dem) 21,853 .4072885 .2660364Women’s Civil liberties (V-Dem) 26,348 .4201775 .2949111Women’s Civil Society Participation (V-Dem) 26,000 .3535175 .2746444Women’s Political Participation (V-Dem) 22,200 .4072169 .3147054Polyarchy (V-Dem) 24,995 .2634506 .2614406Impartial public administration 26,160 -.1091299 1.432628Resource dependence (Miller) 13,529 3.560869 9.716799Ln population (Fariss et al.) 20,828 8.516353 1.687284

ii

02

46

8Pe

rcen

t

4 6 8 10 12Ln GDP p.c. (Maddison)

02

46

8Pe

rcen

t

4 6 8 10 12Ln GDP p.c. (Fariss et al.)

02

46

8Pe

rcen

t

0 .2 .4 .6 .8 1Women political participation index

02

46

8Pe

rcen

t

0 .2 .4 .6 .8 1Women civil society participation index

01

23

45

Perc

ent

0 .2 .4 .6 .8 1Women civil liberties index

02

46

Perc

ent

0 .2 .4 .6 .8 1Female Political Empowerment index

Figure A.1: Histograms over Female Political Empowerment Index and its sub-indices aswell as Ln GDP per capita (from different sources). Y-axes measure percent of observations.

iii

Tab

leA

.2:

Sel

ecte

dro

bust

nes

ste

sts,

usi

ng

Mad

dis

onG

DP

p.c

.dat

a

1-yea

rpanels;

DV:GDP

p.c.growth

5-yea

rpanels;

DV:LnGDP

p.c.

DV

mea

suredin

t+

5t+

1t+

3t+

10

t+

5t+

5t+

5t+

5t+

5t+

5t+

10

t+

5t+

5t+

5b/(t)

b/(t)

b/(t)

b/(t)

b/(t)

b/(t)

b/(t)

b/(t)

b/(t)

b/(t)

b/(t)

b/(t)

Fem

ale

pol.

emp.

1.967**

1.557

2.290**

2.058**

2.234*

1.645

2.409**

1.787*

3.020*

0.123**

0.231***

0.164***

0.440***

0.435***

(2.151)

(1.562)

(2.465)

(2.344)

(1.823)

(1.589)

(2.279)

(1.857)

(1.939)

(2.439)

(2.636)

(3.385)

(3.290)

(3.478)

LnGDP

p.c.

-2.533***

-2.031***

-2.429***

-2.685***

-2.537***

-2.569***

-3.402***

-2.589***

-3.511***

0.888***

0.759***

0.971***

0.972***

0.975***

(-8.268)

(-6.412)

(-7.633)

(-9.308)

(-8.070)

(-8.171)

(-9.155)

(-7.861)

(-8.711)

(63.205)

(27.543)

(27.570)

(43.187)

(8.696)

Polyarchy

-0.103

-0.246

(-0.106)

(-0.238)

Imp.publicadm.

0.079

-0.074

(0.631)

(-0.435)

Resourcedep

.-0.032

-0.031

(-1.210)

(-1.155)

Lnpopulation

-0.226

-0.503*

(-0.868)

(-1.800)

LnGDP

p.c.t−

5-0.082**

0.006

(-2.070)

(0.052)

LnGDP

p.c.t−

10

-0.044

(-1.399)

LnGDP

p.c.t−

15

0.021

(0.752)

Countrydummies

YY

YY

YY

YY

YY

YY

Yea

rdummies

YY

YY

YY

YY

YY

YY

YY

N13362

14036

13697

12540

13172

13362

9721

13021

9535

2762

2600

2367

2762

2630

R2

0.085

0.081

0.085

0.085

0.086

0.085

0.108

0.087

0.109

0.947

0.896

0.952

Instru

men

ts129

129

HansenJ-test

.18

.20

Ar(2)

.13

.18

AR(3)

.22

.61

Notes:

∗p<0.1;∗∗

p<0.05;∗∗

∗p<0.01.Errors

are

clustered

byco

untry.

Models1–12are

estimatedbyOLSandModels13-14bySystem

GMM.

iv

Table A.3: Alternative System GMM specifications

Endogenous regressors FPE FPE and (lagged) Ln GDP p.c.Lags used for instrumentation 3–4 2–3GDP source Fariss et al. Maddison Fariss et al. Maddison

b/(t) b/(t) b/(t) b/(t) b/(t) b/(t) b/(t) b/(t)Female pol. emp. 0.318** 0.098 0.347** 0.418*** 0.280*** 0.281*** 0.143** 0.101**

(2.046) (0.653) (2.339) (2.751) (2.975) (3.002) (2.331) (2.135)Ln GDP p.c. 0.994*** 1.435*** 1.004*** 1.153*** 0.981*** 0.949*** 0.988*** 1.093***

(43.498) (8.280) (44.015) (9.355) (45.490) (9.937) (83.086) (22.363)Ln GDP p.c.t− 5 -0.435** -0.170 0.038 -0.109**

(-2.509) (-1.384) (0.444) (-2.105)Year dummies Y Y Y Y Y Y Y YN 3165 3018 2762 2630 3165 3018 2762 2630Instruments 123 123 126 126 251 329 257 337Hansen J-test 0.527 0.222 0.479 0.000 1.000 1.000 1.000 1.000Ar(2) 0.663 0.283 0.099 0.310 0.497 0.262 0.587 0.295Ar(3) 0.043 0.040 0.845 1.000 0.999 0.058 0.078 0.331

Notes: ∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01. 5-year panel units; DV measured in year t+ 5.

v

Table A.4: Interaction specifications assessing heterogeneity according to time period, region,and regime type

b/(t) b/(t) b/(t)Female political empowerment 2.313** 2.029** 2.965**

(2.258) (2.338) (2.598)Female pol. emp. X post-WWII dummy -0.182

(-0.189)Female pol. emp. X Western countries (dummy) 0.627

(0.787)Female pol. emp. X Democracy (dummy) 0.331

(0.321)Democracy -0.463

(-0.757)Ln GDP p.c. -1.233*** -1.268*** -1.511***

(-4.925) (-4.795) (-5.412)Country dummies Y Y YYear dummies Y Y YN 15879 15879 12760R2 0.029 0.029 0.085

Notes: ∗p<0.1; ∗∗p<0.05; ∗∗∗p<0.01. OLS with errors clustered by country.

vi